Have you ever heard of a “static bubble?” It’s not an easy concept to accept, even to grasp, but it may well be what we have to think about with AI. Is the level of AI investment made by the hyperscaler giants ever going to pay off? Maybe, even probably, eventually. Will these players stop hyping AI, though, blowing the bubble even bigger? Unlikely, even perhaps impossible. Will that eventually blow up in our faces? That’s the “static bubble” question.
The Street has obviously gotten way more concerned about an AI bubble recently, and I’ve been blogging from almost the first warning about an AI hype wave. However, I also pointed out that my AI research suggested that while the big-cloud chatbot LLM version of AI wasn’t exactly setting enterprise CFOs into fits of ecstasy, there were applications of even that AI version that could justify current investment by hyperscalers, and there were other AI models that enterprises did believe would pay off handsomely down the line. Some believe more in AI, of course, and some agree that it’s cloud-AI and not AI that’s a problem.
Nvidia’s earnings report last week initially sent them, and almost the whole tech market up, but that quickly ended in a stock market short-selling frenzy. Clearly, the Street is not totally convinced that AI valuations aren’t still excessive, and some publications also express doubts. I have the same doubts, as I shared with my Wall Street friends right after the call. Things were better by Friday’s close, but on the Monday before Thanksgiving, Nvidia took a hit from a story that Meta and Google were doing a deal for AI chips.
This, to me at least, is a pretty clear cycle of hype and short-selling, indicative of a market that doesn’t have as much confidence as hope, but the balance is close. In the net, AI is hanging fire, in a kind of static bubble. So the question, IMHO, is less whether there’s a bubble or not, but why the bubble seems to be holding on despite all the supposed financial-market concern. I think there are four reasons.
Number one, no surprise, is that Wall Street loves bubbles. You can make money on a market that’s going up. If you’re a professional investor like a hedge fund, you can make money when it goes down too. What you don’t want is a market that isn’t going anywhere at all. So, Wall Street will ride a bubble up, and will try to turn the market by short selling when it’s toppy, which means that they can make money on all the moves bubble-hood creates. They’ll jump on hype when it’s even semi-credible, and bet against it at any whiff of something that can be called bad news.
The second reason is that vendors want to promote a positive view of their own future. If hedge funds are looking for any tiny wound to call it a fatality, and short your stock, then you don’t want to give them anything. In fact, you want to be right there cheering the space you’re in, whatever the truth (as long as you don’t get caught). Many believe that AI vendors, including Nvidia and OpenAI, are actively promoting AI hype. Yeah, it’s probably true; it’s called “marketing”. The need to promote your own position is particularly acute in a hype wave, because if you get sensible when all the rest are still singing market praises, your stock is likely to be hit very hard. For those who’ve read Aleksandr Solzhenitsyn, you don’t want to be the first one to stop clapping.
Reason number three is that news means novelty, not “truth” (for those interested, “pravda” means truth in Russian). An exciting story gets clicked on, and a boring one does not, even if the former one is totally false and the latter totally true. How many sites do you suppose look first at search-engine optimization and second at truth and market reality? In fairness, what else are they supposed to do? I’ve done a lot of sales training and sales management consulting, and I always tell executives that their sales force will do what they’re compensated to do, not what you tell them to do.
Reason four is the only one of the lot that isn’t (perhaps) cynical. Market reality often emerges from hype. Truth is boring, and slow to mature, even to become visible and interesting. If you’re in a hype-driven space like AI, you may well see that there is really going to be something valuable, even transformational, emerge. Wouldn’t it be nice if you could stay alive as a company, if you could keep your job, and enjoy the onrushing commercial reality? So sing pretty now, and be alive if the best is yet to come.
Well, whether that works probably depends on whether you’d prepared, at least a bit, for the reality (or cut and run while the getting was good). So, what is the reality?
That AI is simply software, more sophisticated in architecture, more revolutionary in goals, but software. The software revolution of the past wasn’t one program, it was software overall. Same with AI, and with any technology. There is no single super-solution to all business problems, no single bandage for all wounds. The future of AI is multiplicity, not an LLM or an SLM or ML, but all of these in their own place. What’s happening now is a static bubble, where the forces in balance are the market’s own dispersed initiatives to align AI with real value propositions, and those of the AI giants to ensure that their own massive investments get as much a payback as possible.
But bubbles are more than a battle between surface tension and interior pressure, they’re also dependent on the outside, which in this case is the financial markets. The question for AI, as I’ve been saying, is whether a rational approach, the “dispersed initiatives” and the AI LLM investment can harmonize before financial-market concerns crush everything. In a perfect world, a real market would emerge from competitive efforts to promote one. Today, the problem is that tension between “real” sustainable AI and the massive amount of hype-driven investment in LLM hosting by the giants.
Enterprises, as I’ve noted many times, have always seen AI differently. It’s not a kind of personal coach, it’s a functional element in a workflow. That doesn’t foreclose missions where AI answers questions, it just focuses those missions on where AI can make a significant contribution to a high-labor-value worker. “My company isn’t going to prosper because people write better emails or do it faster,” one CIO told me. “It’s going to prosper if high-level decision-making is better, so you have to focus it on high-level decision-makers and what they need.” Copilot-like technology, aimed at masses of office workers, isn’t moving the ball for this CIO, or for most enterprises who’ve offered me their views.
This is interesting for a lot of reasons, but one that’s rarely discussed is the impact it might have on job loss. Most enterprises (over two-thirds, in fact) tell me that their AI goals are not about cutting jobs at all, and that their most effective AI projects haven’t had that effect. Could this whole AI-steals-job thing be baloney? Maybe so.
I’ve noted before that I think there’s way too much AI hype, and that there is a risk it would turn out to be a true bubble, where hype drives excess expectations and over-investment, but that’s not yet guaranteed. The current level of investment could be justified, but IMHO not by continuing to believe in the hosted-chatbot form of AI, or in “agents” that are really just chatbot retreads. I think that most of the AI giants know the truth and are trying to balance the risk of popping the bubble with nothing to replace it, and the risk of popping with the bubble down the line. They’re all taking different tacks, so we’ll see who, if anyone, gets it right.
